Locality, Delays, and Internal Feedback in Sensorimotor Control Part 1: Motivation and Introductory Theory

Neural architectures in organisms support efficient and robust control that is beyond the capability of engineered architectures. Unraveling the function of such architectures is highly challenging; they contain massive diversity and heterogeneity in components and cryptic internal feedback pathways (IFPs). We introduce the key concepts of speed-accuracy tradeoffs (SATs), diversity-enabled sweet spots (DESS), and controller architecture which aim to help decipher complex neural architecture. As a case study in diversity and complex architectures, we consider nervous system control of movement. In this system, diversity in sensors corresponds to diverse morphology, physiology, and biochemistry of neurons. which can be interpreted as SATs in component specification. While such tradeoffs are often implicit in controller design, we aim here to make them explicit in the canonical LQR setting. We consider an LQR problem with two types of sensors, one fast but sparse and one dense but slow. The sensors achieved substantially lower performance costs when combined than when separate, demonstrating a DESS. The DeSS is also observed for the dual problem in actuation, corresponding to diverse muscle fibers. Controller-internal dynamics, sometimes called internal feedback pathways (IFPs) in biology, were necessary to achieve these combinations. This paper, intended as the first of two parts, bridges familiar control theory problems and novel applications of the System Level Synthesis framework to problems in biology and cyber-physical systems. We additionally outline future experimental plans which will utilize this new theory.